refactor + enable tangent law stereo panning for ISM in the renderer (5d632dd6) · Commits · IVAS Codec Public Collaboration / IVAS Codec

lib_com/ivas_prot.h

+9 −0

Original line number	Diff line number	Diff line
		@@ -4673,6 +4673,15 @@ void ivas_ism_render(
		const int16_t output_frame /* i : output frame length per channel */
		);

		#ifdef FIX_REND_ISM_STEREO_PANNING
		void ivas_ism_get_stereo_gains(
		const float azimuth, /* i : object azimuth */
		const float elevation, /* i : object elevation */
		float left_gain, / o : left channel gain */
		float right_gain / o : right channel gain */
		);

		#endif
		void ivas_mc2sba(
		IVAS_OUTPUT_SETUP hIntSetup, /* i : Format of decoder output */
		float buffer_td[][L_FRAME48k], /* i/o: MC signals (on input) and the HOA3 (on output) */

lib_com/options.h

+1 −0

Original line number	Diff line number	Diff line
		@@ -181,6 +181,7 @@

		#define FIX_REND_ISM_XFADE /* Issue 193: Crossfade inconsistencies in ISM between decoder and external renderer */
		#define FIX_REND_ISM_POS_ROUNDING /* Issue 193: (TEMPORARY! Pending fix for #215) Align rounding of ISM position data in external renderer */
		#define FIX_REND_ISM_STEREO_PANNING /* Issue 193: Use tangent panning for ISM to stereo in external renderer */

		/* ################## End DEVELOPMENT switches ######################### */
		/* clang-format on */

lib_dec/ivas_ism_renderer.c

+48 −0

Original line number	Diff line number	Diff line
		@@ -144,6 +144,12 @@ void ivas_ism_render(
		{
		if ( st_ivas->intern_config == AUDIO_CONFIG_STEREO )
		{
		#ifdef FIX_REND_ISM_STEREO_PANNING
		ivas_ism_get_stereo_gains( st_ivas->hIsmMetaData[i]->azimuth,
		st_ivas->hIsmMetaData[i]->elevation,
		&gains[i][0],
		&gains[i][1] );
		#else
		float aziRad, eleRad;
		float y, mappedX, aziRadMapped, A, A2, A3;
		const float LsAngleRad = 30.0f * PI_OVER_180;
		@@ -174,6 +180,7 @@ void ivas_ism_render(
		gains[i][0] = sqrtf( A3 );
		gains[i][1] = sqrtf( 1.0f - A3 );
		}
		#endif
		}
		else
		{
		@@ -232,3 +239,44 @@ void ivas_ism_render(
		}
		return;
		}

		#ifdef FIX_REND_ISM_STEREO_PANNING
		void ivas_ism_get_stereo_gains(
		const float azimuth, /* i : object azimuth */
		const float elevation, /* i : object elevation */
		float left_gain, / o : left channel gain */
		float right_gain / o : right channel gain */
		)
		{
		float aziRad, eleRad;
		float y, mappedX, aziRadMapped, A, A2, A3;
		const float LsAngleRad = 30.0f * PI_OVER_180;

		/* Convert azi and ele to an azi value of the cone of confusion */
		aziRad = azimuth * PI_OVER_180;
		eleRad = elevation * PI_OVER_180;
		y = ( sinf( aziRad ) * cosf( eleRad ) );
		mappedX = sqrtf( max( 0.0f, 1.0f - ( y * y ) ) );
		aziRadMapped = atan2f( y, mappedX );

		/* Determine the amplitude panning gains */
		if ( aziRadMapped >= LsAngleRad )
		{ /* Left side */
		*left_gain = 1.0f;
		*right_gain = 0.0f;
		}
		else if ( aziRadMapped <= -LsAngleRad )
		{ /* Right side */
		*left_gain = 0.0f;
		*right_gain = 1.0f;
		}
		else /* Tangent panning law */
		{
		A = tanf( aziRadMapped ) / tanf( LsAngleRad );
		A2 = ( A - 1.0f ) / max( 0.001f, A + 1.0f );
		A3 = 1.0f / ( A2 * A2 + 1.0f );
		*left_gain = sqrtf( A3 );
		*right_gain = sqrtf( 1.0f - A3 );
		}
		}
		#endif

lib_rend/lib_rend.c

+35 −17

Original line number	Diff line number	Diff line
		@@ -3791,6 +3791,24 @@ static ivas_error renderIsmToMc(
		wmops_sub_start( "renderIsmToMc" );

		/* TODO(sgi): Possible optimization: less processing needed if position didn't change */
		#ifdef FIX_REND_ISM_STEREO_PANNING
		if ( *ismInput->base.ctx.pOutConfig == IVAS_REND_AUDIO_CONFIG_STEREO )
		{
		set_zero( currentPanGains, 16 );
		ivas_ism_get_stereo_gains( ismInput->currentPos.azimuth,
		ismInput->currentPos.elevation,
		&currentPanGains[0],
		&currentPanGains[1] );

		set_zero( previousPanGains, 16 );
		ivas_ism_get_stereo_gains( ismInput->previousPos.azimuth,
		ismInput->previousPos.elevation,
		&previousPanGains[0],
		&previousPanGains[1] );
		}
		else
		#endif
		{
		#ifdef FIX_REND_ISM_POS_ROUNDING
		// TODO tmu review when #215 is resolved
		if ( ( error = getEfapGains( *ismInput->base.ctx.pEfapOutWrapper,
		@@ -3815,7 +3833,7 @@ static ivas_error renderIsmToMc(
		{
		return error;
		}

		}
		/* Assume num channels in audio buffer to be 1.
		* This should have been validated in IVAS_REND_FeedInputAudio() */
		renderBufferChannelLerp( ismInput->base.inputBuffer, 0, currentPanGains, previousPanGains, outAudio );

tests/renderer/constants.py

+6 −6

Original line number	Diff line number	Diff line
		@@ -376,11 +376,12 @@ pass_snr = {
		"test_ism_binaural_static[ISM2-BINAURAL_ROOM]": 21,
		"test_ism_binaural_static[ISM3-BINAURAL_ROOM]": 21,
		"test_ism_binaural_static[ISM4-BINAURAL_ROOM]": 21,
		# Failure Reason: Casting of positions in renderer to int16_t vs. float in python
		# Failure Reason: Tangent law panning missing in python scripts
		"test_ism[ISM1-STEREO]": 50,
		"test_ism[ISM2-STEREO]": 54,
		"test_ism[ISM4-STEREO]": 55,
		"test_ism[ISM3-STEREO]": 51,
		# Failure Reason: Casting of positions in renderer to int16_t vs. float in python
		"test_ism[ISM1-5_1]": 43,
		"test_ism[ISM1-5_1_2]": 43,
		"test_ism[ISM1-5_1_4]": 43,
		@@ -482,11 +483,6 @@ pass_snr = {
		"test_ism_binaural_static_vs_decoder[ISM3-BINAURAL]": 72,
		"test_ism_binaural_static_vs_decoder[ISM4-BINAURAL]": 71,
		# TODO harmonize panning to stereo
		# Failure reason ISM to stereo panning is done via EFAP in the renderer and tangent law in decoder, harmonize
		"test_ism_vs_decoder[ISM1-STEREO]": 8,
		"test_ism_vs_decoder[ISM2-STEREO]": 17,
		"test_ism_vs_decoder[ISM3-STEREO]": 14,
		"test_ism_vs_decoder[ISM4-STEREO]": 14,
		# Failure reason: Decoder sets elevation for non-planar output layouts to 0
		"test_ism_vs_decoder[ISM1-5_1]": 27,
		"test_ism_vs_decoder[ISM1-7_1]": 27,
		@@ -505,24 +501,28 @@ pass_snr = {
		"test_ism_vs_decoder[ISM1-FOA]": 27,
		"test_ism_vs_decoder[ISM1-HOA2]": 27,
		"test_ism_vs_decoder[ISM1-HOA3]": 27,
		"test_ism_vs_decoder[ISM1-STEREO]": 27,
		"test_ism_vs_decoder[ISM2-5_1_2]": 32,
		"test_ism_vs_decoder[ISM2-5_1_4]": 32,
		"test_ism_vs_decoder[ISM2-7_1_4]": 32,
		"test_ism_vs_decoder[ISM2-FOA]": 32,
		"test_ism_vs_decoder[ISM2-HOA2]": 32,
		"test_ism_vs_decoder[ISM2-HOA3]": 32,
		"test_ism_vs_decoder[ISM2-STEREO]": 32,
		"test_ism_vs_decoder[ISM3-5_1_2]": 33,
		"test_ism_vs_decoder[ISM3-5_1_4]": 33,
		"test_ism_vs_decoder[ISM3-7_1_4]": 33,
		"test_ism_vs_decoder[ISM3-FOA]": 33,
		"test_ism_vs_decoder[ISM3-HOA2]": 33,
		"test_ism_vs_decoder[ISM3-HOA3]": 33,
		"test_ism_vs_decoder[ISM3-STEREO]": 33,
		"test_ism_vs_decoder[ISM4-5_1_2]": 31,
		"test_ism_vs_decoder[ISM4-5_1_4]": 31,
		"test_ism_vs_decoder[ISM4-7_1_4]": 31,
		"test_ism_vs_decoder[ISM4-FOA]": 31,
		"test_ism_vs_decoder[ISM4-HOA2]": 32,
		"test_ism_vs_decoder[ISM4-HOA3]": 31,
		"test_ism_vs_decoder[ISM4-STEREO]": 32,
		# TODO needs investigation
		"test_multichannel_binaural_headrotation_vs_decoder[5_1-BINAURAL-full_circle_in_15s]": 74,
		"test_multichannel_binaural_headrotation_vs_decoder[5_1-BINAURAL-rotate_yaw_pitch_roll1]": 74,